*3.2. Testosterone*

Testosterone (T) signaling is required for spermatogenesis to proceed beyond meiosis. Consequently, it has been postulated that patients with hypogonadism (serum T < 300 ng/dL) may have lower chances of SSR compared to patients with normal serum T levels. Indeed, a pooled estimate of six studies evaluating 2029 patients with NOA undergoing mTESE demonstrated that patients with normal T levels had a significantly higher chance of SSR compared to those with subnormal T levels (OR 1.63, 95% CI 1.08–2.45, *p* = 0.02) [41]. However, the available evidence has provided conflicting results. Reifsnyder et al. evaluated 736 men undergoing mTESE; 348 (47.3%) with baseline T level < 300 ng/dL and 388 (53%) with baseline testosterone levels greater than 300 ng/dL. Among patients with hypogonadism, 88% received hormonal treatment. SRR did not vary among men with low vs. normal baseline T levels; yet, the mean presurgical T level was normal in patients with previous low baseline T levels as the effect of hormonal treatment. Moreover, 18% of patients receiving hormonal treatment did not respond to treatment, but their SRR was comparable to that of responders to treatment [42]. Enatsu et al. evaluated 329 patients, of whom 65 had KS, and found that serum T levels did not differ among men with SSR (97) and SRF (232) (420 + 180 vs. 430 + 190 ng/dL; *p* = 0.42) [43]. Althakafi et al. evaluated 421 patients, of whom 181 had low baseline T levels, and found no difference in SRR between those with normal and low T levels (SRR 38.6% vs. 40.3%, *p* = 0.718). Fifty patients received hormonal treatment with clomiphene citrate (CC) or human chorionic gonadotropin (hCG) due to subnormal T levels: their SRR was comparable to that of patients with normal baseline T levels (36% vs. 38%, *p* = 0.736) [44]. Kizilkan et al. evaluated 860 patients and found that T levels were predictive of SSR in univariate, but not in multivariate, logistic regression [20]. On the other hand, Mehmood et al. and Çayan et al., evaluating 264 and 327 patients respectively, found that SRR was significantly lower in men with low baseline T levels compared to those with normal baseline T levels (40.6 vs. 57.25, *p* = 0.0068, and 40.5% vs. 65.9%, *p* < 0.0001

respectively) [15,45]. Accumulating evidence suggests that higher baseline T levels may be associated with a higher probability of SSR in men with KS [26,34].

It has been suggested that intratesticular testosterone (ITT) measurement could represent a more reliable way of assessing the role of testosterone on the probability of SSR in men with NOA. Due to the inherent risks of performing testicular aspiration to obtain a direct assessment of ITT level, a measurement of the circulating levels of 17-hydroxyprogesterone (17OHP) has been proposed as an indirect biomarker of ITT levels, since 17 OHP is likely to be of testicular and not adrenal origin in men. Indeed, serum 17 OHP levels were found to be undetectable in men receiving exogenous testosterone replacement therapy, and to increase after CC and hCG treatment [46]. Studies evaluating the predictive ability of serum 17 OHP on the probability of SSR in patients with NOA are needed to provide evidence in support or against such a hypothesis.

## **4. Testis Histology**

There is grea<sup>t</sup> consensus about the close relationship between different histopathological categories and mTESE outcome: patients with SCO have the lowest probability of SSR (22.5–41%), while patients with HS have the best chances of sperm retrieval (73–100%), and patients with late MA have better prognosis (SRR 27–86%) compared to those with early MA (SRR 27–40%) [47]. Indeed, a meta-analysis evaluating 19 articles showed that HS predicted SSR (pooled diagnostic odds ratio (DOR) 16.49, 95% CI: 9.63–28.23) with a sensitivity of 30% and specificity of 98%, AUC 0.6758; SCO had a negative predictive ability on SSR (AUC 0.27), while MA had a poor predictive accuracy (AUC 0.55) [25].

To obtain a realistic picture of the severity of spermatogenic dysfunction, the testicular specimen sent to the pathologist should be representative of the overall appearance of the testicular parenchyma. However, it is not uncommon for men with NOA to have more than one histopathological report. Very recently, Punjani et al. demonstrated that these patients may display up to four distinct histopathological subcategories, the increasing histopathological variety being associated with a higher probability of SSR (SRR was 33% in men with one histopathological subtype, compared to 94% in men with 4 subtypes) [48].

Testis histology has been found to be predictive of SRR also in men undergoing salvage mTESE after a failed surgical attempt. Despite previous surgery possibly harming the blood supply of the testis with a potential risk of testicular tissue damage, Tsujimura [49] and Kalsi et al. [50] found comparable SRRs in patients undergoing salvage mTESE after failed cTESE stratified according to testis histology (39% and 40% in SCO, 41.7% and 36% in MA, and 100% and 75% in HS, respectively). Data from Xu et al. [51] confirmed that HS associates with high SRR (85%) even in patients with previous sperm retrieval attempts, but found lower SRR in patients with SCO (5.5%) and MA (25%) compared to previous reports. Very recently, our group found that early and late MA were associated with the lowest probability of SSR (8.7 and 11.1%, respectively), while sperm was retrieved in 85% of men with HS; SRRs in patients with SCO differed significantly according to the presence (focal SCO) or not (complete SCO) of residual areas of HS (SRR 100% vs. 24.4%, respectively) [52].

The obvious limit of testis histology is that it may be obtained only after surgery, therefore it may be used to counsel patients about the probability of having their testicular sperm retrieved in further surgical attempts. In occasional situations, however, testicular histology may be available when a diagnostic testicular biopsy has been done prior to microTESE, and there may be of help in the counselling of patients with NOA.

#### **5. Molecular Markers Expression in the Seminal Plasma**

Given the limited accuracy of hormonal and clinical parameters in predicting the probability of SSR in patients with NOA prior to surgery, researchers have sought to evaluate the feasibility of using the expression of some molecular markers in the seminal plasma as markers of residual spermatogenesis in such patients.

The evaluation of germ cell-specific mRNAs as predictors of SSR in patients with NOA has brought conflicting results. Following the demonstration that the testicular expression of ESX1, an X-linked homeobox gene, was restricted to germ cells, particularly the spermatogonia/preleptotene spermatocytes and round spermatids, and correlated with SSR [53], a group of researchers found that the seminal plasma levels of ESXI were significantly lower in men with NOA compared to normozoospermic subjects (*p* < 0.0001), and predicted SSR in men with NOA with a sensitivity of 84%, but with a specificity of 28% [54]. However, in a further study, the seminal plasma of ESXI was found to be comparable among men with NOA and normozoospermic men [55]; on the other hand, the seminal plasma levels of protamine-1 (PRM1) were found to predict SSR with a sensitivity of 89%, and a specificity of 90%. In another study, however, seminal plasma of PRM1, together with PRM2, DAZ and AKAP4, although being undetectable in patients with SCO, could not predict SSR [56]. Finally, several studies have evaluated the predictive ability of seminal DDX4 mRNA expression on SSR, but again, with conflicting results [reviewed in [57].

Seminal plasma also contains high concentrations of extracellular vesicles that are consistent with exosomes, which originate from the male reproductive tract, and contain coding and noncoding RNAs that vary according to their origin, enabling them to (hypothetically) reflect the pathophysiological conditions of the organ of origin. Some microRNAs (miRNAs) have been found to be preferentially expressed and localized to spermatocytes and spermatids (miR-34b/c and miR-449) or late-stage male germ cells (miR-122), and to be differentially expressed in testis biopsies of patients with and without elongated spermatids (miR-449a, miR-34c-5p and miR-122) [58]. A study evaluated the expression of exosomal miRNAs in the seminal plasma of infertile men with NOA or obstructive azoospermia, demonstrating that three miRNAs, miR-31-5p, miR-539-5p and miR-941, were downregulated in patients with obstructive azoospermia compared to men with NOA. The further evaluation of 12 patients with NOA with (*N* = 8) or without (*N* = 4) SSR showed that the association of the expression values of miR-539-5p and the miR-941 was predictive of SSR [59]. However, due to the very small sample size of such a study, further studies are warranted to provide conclusive results. Indeed, another study found that miR-539-5p was not predictive of SSR, nor could it discriminate normozoospermic, oligozoospermic, and azoospermic men from each other [60].

Long noncoding RNAs (lncRNAs) have been found to play a critical role in spermatogenesis: specifically, they have been implicated in regulating protein-coding genes at the epigenetic level, and it has been speculated that these germ-specific lncRNAs may be involved in epigenetic regulation during spermatogenesis [61]. Many of them display restricted expression in the testis, thus enabling their use as noninvasive biomarkers of spermatogenesis in men with NOA. A recent study investigated the predictive ability of extracellular vesicle long noncoding RNAs (exlncRNAs) in patients with NOA: after having selected 16 exlncRNAs on the basis of their different expression in normozoospermic and azoospermic patients, the Authors evaluated their diagnostic accuracy in predicting SSR in 30 patients with NOA who had (*N* = 18) or not (*N* = 12) their testicular sperm retrieved by mTESE. The Authors built a prediction model based on 9 exlncR-NAs (LOC100505685, SPATA42, CCDC37-DT, GABRG3-AS1, LOC440934, LOC101929088, LOC101929088, LINC00343 and LINC00301) and found that it predicted the probability of SSR with a sensitivity of 88.9% and a specificity of 100%, AUC 0.986. The model was then validated on 66 patients with NOA, with a resulting AUC of 0.960 [60]. Further studies are, however, warranted to validate the findings of the present study, and to confirm or challenge the predictive ability of other molecular markers expressed in the seminal plasma.
